The precise recognition of appropriate synaptic partner neurons is a critical step during neural circuit assembly. However, little is known about the developmental context in which recognition specificity is important to establish synaptic contacts. We show that in the Drosophila visual system, sequential segregation of photoreceptor afferents, reflecting their birth order, lead to differential positioning of their growth cones in the early target region. By combining loss- and gain-of-function analyses we demonstrate that relative differences in the expression of the transcription factor Sequoia regulate R cell growth cone segregation. This initial growth cone positioning is consolidated via cell-adhesion molecule Capricious in R8 axons. Further, we show that the initial growth cone positioning determines synaptic layer selection through proximity-based axon-target interactions. Taken together, we demonstrate that birth order dependent pre-patterning of afferent growth cones is an essential pre-requisite for the identification of synaptic partner neurons during visual map formation in Drosophila.https://doi.org/10.7554/eLife.13715.001
A nervous system requires a precise network of connections between cells called neurons to work properly. Within the brain, the fiber-like connections between pairs of neurons are not running crisscross like a pile of spaghetti. Instead, connected partner neurons are organized into distinct layers and columns.
Many questions remain about how these partner neurons find each other and how the layers of fiber-like connections form. To answer these questions, scientists often study the part of the fruit fly nervous system that controls the insect’s vision. This brain-like structure is simple and can be easily manipulated with genetic engineering. Fruit fly studies have helped identify some molecules that play a role in helping partner cells find one another and connect. These studies have also shown that the timing of brain cell development appears to play a role. But the role that layer formation plays in the process is still a mystery.
Now, Kulkarni et al. show that the birthdate of neurons in the fruit fly visual system helps organize them into layers. These neurons are generated early in the development of the fly. Shortly after birth, a molecular clock under the control of a protein called Sequoia starts within each newly generated neuron. The Sequoia protein is a transcription factor and controls the activity of many genes, and the molecular clock provides the growing neuron fibers with information about where and when to look for its partner neurons.
By manipulating the amount and time that Sequoia is produced in the fly visual system, Kulkarni et al. show that this clock helps arrange the growing cells into layers. Cells with similar birthdates connect and are arranged into layers. How much and when Sequoia is produced dictates where each new layer begins. The next steps for the research will be to learn more about how the clock works and identify any intermediaries between the clock and cell growth patterns.https://doi.org/10.7554/eLife.13715.002
The identification of mechanisms that regulate the precise formation of neural circuits has been one of the major goals in developmental neurobiology. The Chemoaffinity hypothesis, formalized by Roger Sperry (Sperry, 1963), suggests that growing neurons must carry individual identification tags that allow the recognition between synaptic partners. Although many cell-type specific recognition molecules essential for neural circuit assembly have been identified in recent years (reviewed in Missaire and Hindges, 2015; Yogev and Shen, 2014), the precise developmental context in which these molecular tags control cell recognition and specify synaptic identity remain largely elusive.
The temporal pattern in which different types of neurons are generated and specified has been shown to influence their connectivity during further development (Kohwi and Doe, 2013; Osterhout et al., 2014; Pujol-Marti et al., 2012). In addition to cell type specification, all neurons undergo similar steps of cellular differentiation including the growth of specific processes with different molecular and functional properties (Rolls, 2011; Tahirovic and Bradke, 2009) accompanied by the expression of general neuronal molecules like N-Cadherin (Gärtner et al., 2015). These molecules common to most neurons also influence axon targeting (Brusés, 2011; Sakai et al., 2012) and synaptogenesis (Basu et al., 2015; Bekirov et al., 2008; Seong et al., 2015) but how their ubiquitous expression can support neuronal recognition is not well understood.
The Drosophila visual system, due to its highly stereotypic arrangement and genetic tractability, provides an excellent system to understand the mechanisms involved in neural circuit assembly (Clandinin and Zipursky, 2002). Each of the compound eyes is composed of approximately 800 units called ommatidia (Campos-Ortega, 1980) and each ommatidium contains eight photoreceptor or retinula cells (R1-R8). Axons of R1-R6 photoreceptors terminate in the outermost lamina neuropile (Fischbach and Dittrich, 1989). In contrast, R8 and R7 axons project topographically through the lamina and terminate in the medulla (Figure 1 A). This topographic projection leads to the formation of medulla columns that receive input from R7/R8 cells of the same ommatidium. Within the medulla column R8/R7 axons terminate in two different layers, M3 and M6 respectively (Fischbach and Dittrich, 1989), in which they contact their post-synaptic partner neurons (Fischbach and Dittrich, 1989; Gao et al., 2008; Karuppudurai et al., 2014; Melnattur and Lee, 2011; Ting et al., 2014).
The layer specific targeting of R7/R8 axons can be divided into two main developmental phases: First, targeting of R cell axons to distinct temporary layers in the early medulla and second, the selection of correct synaptic target layer within the mature medulla neuropile (Hadjieconomou et al., 2011; Ting and Lee, 2007; Ting et al., 2005). Multiple cell type specific molecules involved in layer specific targeting of R8 axons (Hakeda-Suzuki and Suzuki, 2014; Lee et al., 2003; Ohler et al., 2011; Pappu et al., 2011; Senti et al., 2003 Shinza-Kameda et al., 2006; Timofeev et al., 2012; Tomasi et al., 2008) and R7 axons (Astigarraga et al., 2010a; 2010b; Choe et al., 2006; Lee et al., 2001; Morey et al., 2008; Nern et al., 2005; Prakash et al., 2010; Ting et al., 2005) have been identified. Interestingly, most of these molecules function in the second phase of axon targeting and the molecular mechanisms governing the initial innervation of R8 and R7 axons as well as the importance of this temporary positioning for subsequent synaptic layer targeting remain elusive. The Zn finger transcription factor Sequoia and the cell adhesion molecule N-Cadherin are both expressed in R7 as well as R8 cells but are primarily required for the temporary layer positioning of R7 axons (Lee et al., 2001; Petrovic and Hummel, 2008; Ting et al., 2005). Additionally, the LRR molecule Capricious, expressed only in R8 cells, has been described to control R8 axon targeting during the second step (Shinza-Kameda et al., 2006) but its role in the initial targeting of R8 axons to temporary medulla position has not been addressed despite an early onset of expression.
Here we show that early self-patterning of R7/R8 afferents, mediated by relative difference in Sequoia levels, organizes an initial topographic map. This is achieved by a birth-order defined sequence of R7-R8 growth cone segregation leading to their differential positioning in the target field. Shortly afterwards, cell adhesion molecules, like Capricious in R8, consolidate these growth cone positions, which is critical for subsequent steps of synaptic partner recognition.
In the developing Drosophila visual system, photoreceptor axons project from the eye disc into the optic lobe and target to the lamina and medulla neuropiles (Figure 1A). Photoreceptor differentiation begins in the 3rd instar eye disc in a defined sequential fashion with R8 specified first followed by the outer R1-R6 cells and finally R7 in every ommatidium and can be visualized in developing ommatidial rows (Figure 1B, Tomlinson and Ready, 1987).
We examined how R8/R7 sequential specification is represented in the arrival of R cell axons at the medulla target region using cell type specific reporter lines (sens-Gal4 for R8, sev-Gal4 for R7, Figure 1C–E”, Figure 1—figure supplement 1). By quantifying the number of R8/R7 growth cones at three consecutive stages of early pupal development (6/12/24 hr After Puparium Formation, APF), we could show that R cell axon innervation in the medulla mirrors the temporal pattern of R cell specification and follows a consistent sequence of growth cone segregation for each stage (Figure 1C–E’, summarized in B). At the anterior edge of the medulla the youngest R8 axons arrive sequentially and position at the superficial layer of the medulla neuropile ('R8 positioning', Figure 1C arrowhead), followed by the arrival of the first R7 axon at the respective R8 position about 6 hr later and three columns posterior to the youngest R8 axon (Figure 1C’ arrowhead). For the next 6-8 hr, represented by 3–4 columns, R7/R8 growth cones are in close contact followed by their segregation into adjacent positions, with R7 growth cones locating proximal with respect to the superficial R8 growth cones in the corresponding columns ('R7-R8 segregation', Figure 1C’-E’). Afterwards, R7 growth cones move deeper into the medulla neuropile whilst maintaining their columnar topography and a separate R7 temporary layer becomes visible ('R7 positioning', illustrated in Figure 1B).
As reported previously, Sequoia is critical for R cell target layer selection (Petrovic and Hummel, 2008). Next we tested if Sequoia is involved in the regulation of these initial steps of growth cone positioning. Using MARCM (Lee and Luo, 2001) we generated sequoia mutant R8 cells by activating Flippase under heat shock promoter and visualized their growth cones with an R8 specific reporter line (sens-Gal4, Figure 2A–B’). Axons of sequoia mutant R8 cells project to the anterior medulla in the wild type sequence but convergence of neighbouring R8 growth cones can be observed, disrupting the topographic organization (Figure 2A–B'). Similar phenotype in R8 growth cone positioning can be observed in eye specific sequoia mosaics (ey3.5-FLP), excluding an effect of unlabelled sequoia mutant cells generated in the brain in the hs-FLP background (data not shown).
Similar to R8, no defects can be detected in the extension and arrival of sequoia mutant R7 axons but they fail to segregate from R8 growth cones within the same column as early as 6 hr APF (Figure 2C–D'). Subsequently, these sequoia mutant R7 growth cones remain at the superficial medulla position together with R8 growth cones even at 24 hr APF (Figure 2E–F'). Interestingly, sequoia mutant R7 growth cones from two neighbouring columns also converge in the superficial medulla position (Figure 2G–G'). Single cell analysis revealed a cell-autonomous function of Sequoia in growth cone segregation (Figure 2—figure supplement 1A–B', arrowhead). Therefore, loss of Sequoia function disrupts the sequential segregation of R8-R8, R8-R7 and R7-R7 growth cones within as well as between layers.
The targeting phenotype of sequoia mutant R7 cells varies depending on the genetic background of R8 in the same medulla column: In case of wild type R8 axons, sequoia mutant R7 from the same ommatidium will terminate together in M3 (Figure 2I,I'), whereas a sequoia mutant R8 and R7 co-terminate in M1 (Petrovic and Hummel, 2008). To test if mis-targeting of sequoia mutant R7 cell axons is caused by a change in their target layer recognition or the consequence of an afferent segregation defect, we analysed the phenotype of wild type and sequoia mutant R7 axons in a background with a modified R8 axon position. R8 axons were retained in the superficial medulla layer M1 via Golden-goal (gogo) over-expression using GMR-gogo, which has no effect on the targeting of R7 axons (Tomasi et al., 2008). In this background, we labelled wild type and sequoia mutant R7 cells using MARCM and analysed axon targeting in the adult medulla. Wild type R7 growth cones segregate from ectopic R8 growth cones and reach the layer M6 (Figure 2J–J'), indicating that the ectopic position of R8 axons does not influence wild type R7 axon targeting. In contrast, sequoia mutant R7 axons mis-target to layer M1 along with ectopic R8 axons (Figure 2K–K'). This indicates that mis-targeting of sequoia mutant R7 axons is the consequence of growth cone segregation defects rather than a change in target layer recognition.
In summary, Sequoia regulates two main steps of growth cone segregation of R7/R8 axons. For axons of the same R cell type but from different ommatidia, Sequoia supports point-to-point spacing within the temporary layer and thereby controls the subsequent columnar restriction. In addition, for the pair of R7/R8 cells from a single ommatidium, which innervate the same medulla column, Sequoia mediates the segregation of growth cones between layers.
Sequoia shows a highly restricted expression during early differentiation of all R cells (Petrovic and Hummel, 2008). A short peak of high Sequoia expression at the onset of R cell differentiation is followed by a rapid decline in protein levels (Figure 1—figure supplement 1A graph). Due to the sequential development of ommatidia in the eye field as well as R cell types within each ommatidium (Tomlinson and Ready, 1987), the short peak of Sequoia expression in each cell leads to small differences of Sequoia levels in sequentially projecting R8 cells from adjacent ommatidial rows ('inter-ommatidial differences', Figure 1—figure supplement 1A, Figure 3K1a). In addition, each R7 develops approximately 8 hr after R8 in every ommatidium corresponding to four ommatidial rows in the eye disc. Therefore, the highest difference in the expression levels of Sequoia can be found between these two cell types with almost no detectable Sequoia in R8 at the time of maximal Sequoia expression in the R7 of the same ommatidium ('intra-ommatidial differences', Figure 3—figure supplement 1C, inset 3, Figure 3K1b).
To determine the role of Sequoia expression dynamics in R7/R8 growth cone positioning we next analysed how inter- and intra-ommatidial differences in Sequoia levels influence growth cone segregation. Abolishing the inter-ommatidial differences via prolonged Sequoia expression using LGMR-Gal4 leads to a convergence of R7/R8 growth cones in the deeper medulla position (Figure 3—figure supplement 1A–B'). Similarly, equalizing Sequoia levels in only R7 cells using sev-Gal4, which shows transient, low-level expression as compared to LGMR-Gal4, resulted in frequent R7 growth cone convergence and corresponding innervation gaps in the deeper medulla layer (Figure 3A–B', arrowheads in B’). This result shows that the loss and gain of Sequoia activity leads to similar growth cone convergence phenotypes indicating that the presence of Sequoia expression is not sufficient to mediate growth cone segregation.
To determine if the observed differences in Sequoia levels between R cells are critical for their growth cone segregation into different layers, we sought to model the endogenous Sequoia expression difference between R8 and R7 cells among neighbouring R7 cells targeting within a layer. For this, we analysed the axon projections in Sequoia gain-of-function R7 cell mosaics (Seqhigh) at the clonal boundary with adjacent wild type R7 cells (Seqlow; Figure 3C–F). Here we observed that the majority of single Seqhigh R7 axons segregate from the surrounding wild type Seqlow axons and project deeper into the medulla (Figure 3D'). In contrast to the single cell clones, most of the mosaics with two adjacent Seqhigh R7 cells show growth cone convergence and termination at the appropriate medulla position (Figure 3D”). Interestingly, single Seqhigh R7 cells from distant columns, which overshoot the temporary layer, often converge in the inner medulla neuropile (Figure 3D' arrow) or even within the lobula complex (data not shown). Similar to R7, mosaics of R8 cells with different Sequoia expression levels segregate during initial axon positioning: single Seqhigh R8 axons segregate from the surrounding wild type Seqlow R8 axons (Figure 3E' arrowhead) and extend further into the medulla neuropile (Figure 3E' arrow), whereas two or more neighbouring Seqhigh R8 growth cones converge and terminate at the R7 position (Figure 3E" arrow). Thus, together with the loss-of-function, this gain-of-function analyses show that ‘equalized’ Sequoia levels prevent growth cone segregation thereby highlighting the role of relative Sequoia levels in this process (summarized in Figure 3K2).
To test if relative differences in Sequoia levels also control growth cone patterning of other photoreceptors, we analysed the outer R1-R6 cells. These neurons develop in a temporal window between R8 and R7 specification (Tomlinson and Ready, 1987) and their growth cones are positioned in a single temporary layer, the lamina plexus, which corresponds to their small relative differences in Sequoia expression (Figure 3—figure supplement 1C, Figure 3G,G'), We generated sequoia mutant clones of R1-R6 cells using ey3.5-FLP and visualized the growth cones of R2/R5 cells (ro-τ-LacZ) or R4 cells (m∂0.5-Gal4) (see Materials and methods). The sequoia mutant R1-6 growth cones fail to segregate normally, illustrated by gaps within the lamina plexus (Figure 3H–H”, Figure 3—figure supplement 1E-E’), which is similar to the defects in inter-ommatidial R8 as well as R7 growth cone segregation. Strikingly, clones of Seqhigh R1-6 cells segregate their growth cones from the wild type (Seqlow) ones thereby organizing the formation of a novel layer within the lamina neuropile (Figure 3I–J”, Figure 3—figure supplement 1F,F’). This demonstrates that large differences in Sequoia levels among R cells, either inherently present as in case of developing R7/R8 cells or ectopically generated as in case of R1-R6 cells, segregate their growth cones into distinct layers. In all cases, cells with higher Sequoia levels position their growth cones deeper in the neuropile. From these data we propose a 'Growth Cone Segregation' model, in which small differences in Sequoia levels lead to an evenly–spaced growth cone positioning within a layer, whereas large Sequoia differences result in segregation of growth cones into separate layers. In contrast, equal levels of Sequoia in projecting R cells, either in loss- or gain-of-function context, cause the convergence of growth cones and termination of axonal extension (Figure 3K).
We next determined the temporal dynamics of initial R8/R7 growth cone segregation into superficial vs. deeper medulla layers. To temporally restrict the Gal4-driven expression of Sequoia, we utilized the Gal80ts (TARGET system, McGuire, 2003). At the permissive temperature (18°C) functional Gal80 prevents Sequoia expression whereas an inactive Gal80 at the restrictive temperature (29°C) allows the induction of Sequoia expression. Using this method, an R8 growth cone shift to deeper medulla position can be observed following Sequoia expression until 24 hr APF (Figure 4—figure supplement 1C). In contrast, an onset of Sequoia expression after 24 hr APF, at which time all photoreceptor axons have arrived in the target field, has no effect on R7/ R8 growth cone segregation (Figure 4—figure supplement 1D). This indicates that R7/R8 cells are sensitive to Sequoia-induced growth cone segregation only during a narrow developmental window. Interestingly, although R8 growth cones do not change the layer upon Sequoia expression after 24 hr APF, they still respond to these elevated Sequoia levels by leaving their topographic position and converging onto neighbouring growth cones (Figure 4—figure supplement 1D). This indicates that the restriction of R cell growth cones to distinct temporary layers occurs immediately following their segregation whereas the process of columnar restriction continues into the later steps of visual map formation (Ferguson et al., 2009; Ting et al., 2007).
To further characterize the developmental sequence of the transition from initial positioning to the consolidation of R8 growth cones, we induced short pulses of Sequoia expression in early pupal stages using the Gal80ts method described above and analysed R8 growth cones at 24 hr APF (Figure 4A–F). Upon the induction of Sequoia expression at 6 hr APF with about 18 R8 growth cones in the medulla target region, we observed a defined shift of the youngest 10 of these R8 growth cones to the deeper R7 position whereas the more posterior, and therefore older, 8 R8 growth cones remain at their superficial medulla position. This indicates that posterior R8 growth cones had consolidated their position prior to the effects of induced Sequoia expression (Figure 4B). Two hours later (at 8 hr APF), four more R8 growth cones (12 out of 20) are no longer responsive to elevated Sequoia levels as they do not leave their superficial medulla positions (Figure 4C). Upon the onset of Sequoia expression at 12 hr APF, 16 out of 24 R8 growth cones are retained in the superficial medulla position (Figure 4D). Additionally, continuous Sequoia expression at later developmental stages does not disrupt the superficial layer positioning of R8 growth cones that are consolidated prior to Sequoia induction (Figure 4E). Together, the consolidation of each R8 growth cone occurs approximately 18 hr after its arrival at the superficial medulla layer (Figure 4F). The fixed correlation between the number of consolidated R8 growth cones and developmental time indicates that for each R8 growth cone there is a defined transition from the initial positioning to the consolidation shortly after R7-R8 segregation.
To gain further insights into the molecular mechanism underlying the consolidation of R8 cell growth cones we tested candidate molecules expressed during initial axon targeting. Capricious (Caps) is expressed in projecting R8 cells and has been proposed to mediate R8 axon targeting (Shinza-Kameda et al., 2006). We determined the relative levels of Capricious expression in the developing medulla at the R8 and R7 positions by measuring the ratio of normalized fluorescence intensities (NFI) of Capricious staining at the R8 and R7 growth cones against the surrounding medulla region (Figure 5 Table). At the anterior medulla, Capricious shows a homogeneous expression with NFI ratios for R cell growth cones and corresponding medulla region being close to 1 (0.94 in R8 and 0.87 in R7 growth cone position, Figure 5A–D). Following the phase of growth cone segregation, Capricious enriches during the process of consolidation at the region of R8 growth cones with an NFI ratio increasing to 1.39 and a four fold decline at the R7 position (with the ratio of 0.24, Figure 5A’–D’). This indicates that during the process of consolidation, Capricious levels in R8 growth cones increase. At the same time, Capricious-negative R7 growth cones move to a deeper medulla position devoid of any Capricious expression.
To determine the role of Capricious in the process of growth cone consolidation we followed the projection of capricious mutant R8 cells using MARCM. Interestingly, no defect could be detected in the initial position of R8 growth cones as well as subsequent R7/R8 segregation (Figure 5E–E”,G,G’). R8 growth cone positioning defects first appear during the phase of consolidation where capricious mutant growth cones extend towards the deeper R7 position (Figure 5G”). Later in development, these capricious mutant R8 axons mis-target to layer M6 as previously reported (Figure 5F–H", Shinza-Kameda et al., 2006). This shows that Capricious is not required for the initial positioning of R8 growth cones and R7/R8 segregation but is critical for the subsequent step of R8 growth cone consolidation in the superficial medulla.
To further elucidate the role of Capricious in the consolidation of R8 growth cones we modified Capricious levels in R8 cells with prolonged Sequoia expression. The partial reduction of Capricious levels via targeted RNAi does not interfere with the consolidation of R8 growth cones in the superficial medulla position (Figure 5I). As described above, approximately 16 R8 growth cones in the posterior medulla region are consolidated in their initial position at 12 hr APF, making them insensitive to the induction of Sequoia expression (Figure 5J). In contrast, the RNAi-mediated Capricious reduction in the background of elevated Sequoia expression severely affected the R8 growth cone consolidation resulting in most of the posterior R8 growth cones shifting to the deeper R7 position (Figure 5K,L). Furthermore, co-expression of Sequoia and Capricious prevents Sequoia-induced shift of R8 growth cones and leads to segregation of R7 and R8 growth cones in a wild type pattern (Figure 5—figure supplement 1A–B’). Therefore, Sequoia mediated growth cone segregation and Capricious mediated adhesion are antagonistic forces that are capable of balancing each other to achieve proper positioning of R-cell growth cones in the developing medulla.
Following the sequential positioning of R7/R8 growth cones during the first half of pupal development, all R7 and R8 axons simultaneously extend during the second half of medulla circuit assembly to the M6 and M3 layer respectively (Ting and Lee, 2007; Özel et al., 2015). To determine how the initial growth cone positioning influences subsequent steps of synaptic layer selection we followed the development of R8 growth cones which have been displaced to deeper R7 position. Wild type R8 growth cones are positioned at the superficial medulla at 24 hr APF (Figure 6A,A’) and later target to layer M3 in the adult (Figure 6B,B’). A pulse of Sequoia expression until 24 hr APF and subsequent repression using Gal80ts leads to an initial shift of all R8 growth cones to the deeper R7 position (Figure 6C,C’). Interestingly, in the subsequent steps of synaptic layer targeting, these R8 axons, together with R7 axons, terminate in the layer M6 even without further Sequoia expression (Figure 6D,D’). Additionally, an early Sequoia pulse until 6 hr APF leads to a shift of 8–10 posterior R8 growth cones towards the deeper R7 medulla position (Figure 6E,E’). Following this pattern during the subsequent steps of medulla development reveals that R8 axons recognize their final target layer exactly according to their initial growth cone position (Figure 6F,F’), with 10 most posterior R8 axons targeting to the layer M6 (Figure 6F’ arrowhead) and the remaining R8 axons terminating in the layer M3 (Figure 6F’ arrow). We analysed if this R8 mis-targeting to layer M6 results from Sequoia-induced changes in the expression of Capricious and the known R8 guidance receptor Frazzled (Pecot et al., 2014; Shinza-Kameda et al., 2006; Timofeev et al., 2012). However, no difference in Capricious and Frazzled expression can be detected in R8 cells with elevated Sequoia levels (Figure 6G-J'), which is in line with earlier data showing that Sequoia does not influence cell-type specific differentiation programs of R7 and R8 (Petrovic and Hummel, 2008).
Interestingly, the main R7 synaptic target cell, Dm8, is already in close proximity of R7 growth cones immediately following the R7/R8 segregation and positioning of R7 growth cones in the deeper medulla (Figure 6—figure supplement 1A–A'), Ting et al., 2014. Therefore the Dm8 neurites and mis-positioned R8 growth cones are also in close proximity raising the possibility that these two cell types can directly interact with each other (Figure 6—figure supplement 1B–B'). To test if the ectopic R8 axons form synapses with Dm8 neurites in layer M6, we made use of the recently modified GFP reconstitution method (syb-GRASP; Karuppudurai et al., 2014). In this method, the GFP1-10 fragment is fused to the C-terminus of Drosophila n-synaptobrevin (n-syb) producing n-syb::spGFP1-10 chimera, resulting in reconstitution with GFP11 only after vesicle fusion (Macpherson et al., 2015) and a preferential labelling of ‘active-synapses’ rather than neuronal contacts made at any time during development. In wild type (Figure 6K–K”), the expression of pre-synaptic syb::spGFP1-10 in R7/R8 (LGMR-Gal4) and post-synaptic LexAop-spGFP11::CD4, in Dm8 cells (OrtC1-3 LexA DBD, OrtC2B dVP16AD; Karuppudurai et al., 2014; See Materials and methods for details), results in GRASP-positive R7->Dm8 connections in layer M6 (Figure 6K” arrows) and no GRASP-positive R8->Dm8 connections can be detected in layer M3 (Figure 6K” arrowheads). When R7 growth cones are retained at the R8 temporary layer via UAS-CapriciousID expression using sev-Gal4 (Shinza-Kameda et al., 2006), GRASP signal, demonstrating R7->Dm8 contacts, can be observed only in columns where 'escaper' R7 axons target to layer M6 (Figure 6L” arrows), but not in columns with mis-targeted R7 axons (Figure 6L” arrowhead). Surprisingly, GRASP-positive R8->Dm8, similar to R7->Dm8, connections can be detected in medulla layer M6 following the early shift of R8 growth cones to the deeper medulla position (Figure 6M–M”), indicating that both R7 and R8 can recognize Dm8 processes and form synaptic contacts.
Here we demonstrate that the early, birth order dependent, segregation of R cell growth cones determines later synaptic layer identity in the Drosophila visual system. Small inter-ommatidial differences in Sequoia levels organize R cell growth cones within a layer whereas large intra-ommatidial differences segregate growth cones between layers. Changes in the positioning of growth cones directly correlate with changes in synaptic layer selection without affecting the expression of known cell-type specific targeting molecules. These results highlight the importance of initial afferent growth cone positioning for visual map formation prior to synaptic partner recognition (Figure 7).
An early shift in the R8 growth cones to R7 position, induced by a short pulse of Sequoia expression, allows them to recognize the R7 target cell Dm8 as synaptic partner later during the development. Similarly, if R7 growth cones fail to segregate from R8 growth cones, they terminate together with R8 axons in the M3 synaptic layer independent of their intrinsic differentiation and targeting program. This extension of Frazzled-negative R7 axons (Timofeev et al., 2012) towards layer M3 could be explained by the default setting of R7 axons to tightly fasciculate and follow the R8 pioneer axons towards the target region (Fischbach and Heisinger, 2008; Meinertzhagen and Hanson, 1993).
R cell growth cone segregation can be controlled by axon-target interactions or axon-axon interactions or both. Although there is no experimental demonstration for direct R7-R8 afferent interactions, the following set of data indicate that such interactions occur during development. Support for direct interaction between R7-R8 afferents comes from Maurel-Zaffran et al., 2001, where the expression of LAR as membrane tethered ligand in R8 cells alone (using an R8-specific Gal4 driver line) could induce a response from R7 axons, indicating a direct signalling between R8-R7 axons. Further, induced Capricious expression in R8 and R7 cells, in Capricious null background (therefore resulting in a target region without any Capricious expression) is sufficient to mis-target R7 axons to layer M3, again indicating a direct R7-R8 afferent interaction (Berger-Müller et al., 2013). Here we show that final target layer of sequoia mutant R7 axons depends on the targeting of R8 axons, further suggesting that mis-targeting of sequoia mutant R7 axons to ectopic synaptic layer is the consequence of segregation defect rather than a change in target layer recognition.
Interactions among afferent axons have been implicated in the assembly of visual and olfactory circuits in vertebrates as well as invertebrates (Brown et al., 2000; Clandinin and Zipursky, 2002; Ebrahimi and Chess, 2000; Feinstein and Mombaerts, 2004; Komiyama et al., 2007; Petrovic and Schmucker, 2015; Sweeney et al., 2007). It has recently been shown that Eph-Ephrin signalling mediates local sorting of RGC axons in mammalian visual system (Brown et al., 2000; Suetterlin and Drescher, 2014). Notch signalling was demonstrated to play a role in spacing of DCN cluster neuron axons via neighbour axon interactions (Langen et al., 2013). But, whether these afferent interactions influence synaptic partner recognition is not known.
Here we show that in the Drosophila visual system, relative levels of Sequoia determine the segregation of afferent R7/R8 growth cones within or between layers. By creating Seqhigh-Seqlow R cell combinations using Sequoia gain-of-function R7 mosaics we observed that difference in Sequoia levels among neighbouring cells could induce growth cone segregation. The endogenous differences in Sequoia levels most likely arise as a result of the temporal sequence of R cell specification, suggesting a self-patterning mechanism in early visual circuit assembly (Hassan and Hiesinger, 2015; Roignant and Treisman, 2009, Tomlinson and Ready, 1987). How the relative differences in Sequoia levels in the nuclei of R cells translate into growth cone segregation remains elusive. We have tested candidate signalling pathways including Semaphorin/Plexin (Cafferty et al., 2006; Hsieh et al., 2014; Pecot et al., 2013; Yu et al., 2010), TGF-beta ligand Activin and its receptor Baboon (Ting et al., 2007) and Notch (Langen et al, 2013) but did not find evidence for a critical role in initial growth cone segregation (data not shown). This suggests a so far unknown molecular mechanism in which the growth state of an axon is directly coupled to differential growth cone adhesion. As we could demonstrate a cell-autonomous function of Sequoia in R8 for columnar segregation as well as in R7 for layer segregation, we envision a mechanistic model related to the concept of cell competition, in which strong cell-cell interactions induce cell-autonomous responses (Rhiner et al., 2010).
The initial segregation of afferent growth cones into distinct positions is then consolidated by expression of Capricious in R8 axons in the same posterior-to-anterior pattern in which they arrive in the medulla. We speculate that Capricious mediated growth cone consolidation serves two purposes: 1. It removes the temporal difference in the arrival of R8 axons and 2. It maintains R8 axons in the position where they are responsive to subsequent NetrinB signal provided by L3 neurites (Pecot et al., 2014; Timofeev et al., 2012). This is supported by two different sets of results: First, as we show in this study, the displacement of R8 growth cones to deeper medulla position leads to their mis-targeting to layer M6 in spite of normal Frazzled expression in these R8 cells. Second, the ectopic expression of Frazzled in R7 cells cannot re-direct them to M3 layer in response to localized NetrinB signal present in the superficial position but R8 axons can be re-directed to a different layer (M1/M2) by ectopic expression of localized NetrinB in a position deeper to the superficial R8 medulla position (Nern et al., 2008; Pecot et al., 2013; Timofeev et al., 2012). Taken together, these observations suggest that M3 layer targeting via L3-mediated NetrinB signalling requires R8 axons to be positioned superficially in the medulla further underscoring the importance of R8 growth cone consolidation in this position (Figure 7).
Previous studies have identified several molecules necessary for M6 targeting of R7 axons (Clandinin et al., 2001; Hofmeyer and Treisman, 2009; Hofmeyer et al., 2006; Maurel-Zaffran et al., 2001; Ting et al., 2005; Tong et al., 2011), including Liprin-alpha/beta/gamma, PTP69D and D-Lar. The loss of these molecules specifically affects the stabilization of R7 growth cones during the second step of targeting. Additionally, these molecules along with N-Cadherin have been shown to be critical for establishment of synaptic contacts between pre and post-synaptic neurons (Arikkath and Reichardt, 2008; Astigarraga et al., 2010b; Garrity et al., 1999; Hummel and Zipursky, 2004; Lee and Godenschwege, 2015; Nagaoka et al., 2014; Prakash et al., 2010; Reines et al., 2012). We confirm previous observations, that R7 growth cones are in close proximity with their primary post-synaptic target neurons, Dm8, at the end of growth cone segregation (Figure 6—figure supplement 1A–B', Ting et al., 2014). This raises the possibility that targeting of R7 axons to M6 layer, later in the development, could be the direct result of R7->Dm8 contacts mediated by N-Cadherin. Recently it was shown that N-Cadherin function is necessary for stabilizing R7 growth cones in the deeper medulla position but not for targeting and subsequent extension of R7 axons to M6 layer seems to be a result of passive dislocation (Özel et al., 2015). Additional support for the role of N-Cadherin in the formation and maintenance of R7->Dm8 contacts, following their initial segregation from R8 growth cones, comes from our observation that early expression of Sequoia in CadN mutant R7 cells under weak elav-Gal4 driver can rescue the mis-targeting of R7 growth cones in the superficial medulla position along with R8 growth cones at 24 hr APF (Figure 6—figure supplement 2A,A’,C,C’), but fails to rescue the later mis-targeting to layer M3 eventually resulting in a mis-targeting phenotype identical to CadN mutant R7 axons (Figure 6—figure supplement 2B,B’,D,D’). Interestingly, the R8 growth cones initially mis-positioned in the deeper medulla eventually mis-target to layer M6 and form synaptic contacts with Dm8. In addition, these R8 cells, with axons mis-targeted to layer M6, do not show changes in any of their known cell-type specific molecules including early specifier of cell identity (Senseless), guidance receptors (Frazzled, Capricious) and sensory receptors (Rh6). In addition, no expression of R7 specific molecules (Prospero, R3, Rh4) can be detected. Thus, the R8 cells interact with Dm8 neurons most likely via ubiquitously expressed molecules such as N-Cadherin expressed in both, R7 as well as R8, cells. This is supported by our observation that N-Cadherin is required for stabilization of R8 axons at the layer M6 (Figure 6—figure supplement 2E–F’).
We observed that R8 axons form functional synapses with Dm8, a known R7 target neuron, in the layer M6. This raises the fundamental question of how synaptic layer selection influences synaptic partner recognition. The cellular complexity of potential post-synaptic target layer encountered by ingrowing R cell axons has not been fully determined, leaving room for selective recognition for synaptogenesis within a layer. In fact, it has been shown that within M6, R7 axons form synapses with Dm8 but not with Tm5c (Karuppudurai et al., 2014) which also arborize the M6 layer. In addition, we have identified various medulla columnar neurons within M6 that are not contacted by R7 axons (unpublished results). Similarly in layer M3 R8 and L3 select distinct post-synaptic partners (Takemura et al., 2013).
The types of neurons present in the medulla at the time of R8 and R7 axon innervation have not been fully identified. Based on the data from Hasegawa et al., 2011; Li et al., 2013; Suzuki et al., 2013, the medulla neurons are generated in temporal fashion (reviewed in Sato et al., 2013; Suzuki and Sato, 2014) and therefore they likely innervate the medulla at different time points. Experiments presented here support a developmental scenario in which the medulla context for arriving R cell axons reduces the complexity of synaptic partner selection. For example R8 and L3 have different arrival times at M3 (Nern et al., 2005; Pecot et al., 2013; 2014), thereby would encounter a different local environment of potential post-synaptic partners competent for synaptogenesis. It is plausible that some form of temporal co-ordination of afferent axons and their post-synaptic partner cell neurites would actually simplify the synaptic partner matching. The concept of temporal identity would argue that R7 and R8 axons arriving at the same medulla position approximately the same time, as shown in the Sequoia gain-of-function background, will pick the same synaptic partners exemplified by Dm8. Support for such proximity-based axon-target interaction for synaptogenesis comes from earlier analysis of ectopic axons in Drosophila as well as Zebrafish (Edwards and Meinertzhagen, 2009; Berger-Müller et al., 2013; Pujol-Martí et al., 2014).
From an evolutionary perspective, such proximity-induced synapse formation has several advantages over mechanisms that require regulation and expression of distinct sets of cell recognition molecules. Considering R7 as the most recently added cell to the precursor ommatidium (Mavromatakis and Tomlinson, 2012): During development, R7 is recruited using mechanisms similar to R8 and therefore possesses default R8 specification program. However, this default R8 program is suppressed to facilitate R7 specification (Cook et al., 2003; Morey et al., 2008). Thus, a temporally separated, novel R7 cell is generated with basic neuronal differentiation similar to that of an R8 cell (Brennan and Moses, 2000; Friedrich et al., 2011; Roignant and Treisman, 2009). Interestingly the temporal difference in the R8/R7 differentiation is then translated into Sequoia mediated layer segregation of their growth cones (Petrovic and Hummel, 2008), with Sequoia expression being part of common differentiation program. Thus, the evolutionary recent R7 cell seems to recognize its synaptic targets via pan neuronal molecules like N-Cadherin as part of the default neuronal differentiation program, instead of the invention of an additional recognition code.
The flies were raised at 25°C unless otherwise mentioned.
The following flies were used in this study:
CantonS, FRT42B, FRT2A, Gal80/II, Gal80/III, FRT80, GMR-GFP, UAS-mCD8GFP, hs-FLP, ey3.5-FLP, GMR-FLP, tub-Gal80ts, elav-Gal4, sev-Gal4, LGMR-Gal4, m∂0.5-Gal4, ro-tau-LacZ, Rh6-EGFP, PanR7-Gal4 were obtained from Bloomington Drosophila Stock Center. sens-Gal4/CyO and sens-Gal4/TM6 were gift from Bassem Hassan. sequoia5 is a previously generated Sequoia loss-of-function allele (Petrovic and Hummel, 2008). UAS-Sequoia was obtained from Jay Brenman. UAS-Capricious, UAS-CapriciousID(intracellular deletion), CapriciousC18fs FRT2A and caps-LacZnls flies were kindly provided by Akinao Nose. GMR-gogo was a generous gift from Takashi Suzuki. PanR8-Gal4 was a gift from Claude Desplan. UAS-CapriciousRNAi was obtained from VDRC (VDRC Transformant ID No. 27097). PM181-Gal4 and CadN405FRT40 (Lee et al., 2001) were used in Sequoia analysis, Early Dm8 labelling OK371-VP16AD/CyO; ortC2-Gal4DBD/TM2, adult Dm8 specific OrtC1-3 LexA DBD, OrtC2B dVP16AD/CyO (Ting et al., 2014) and syn-GRASP constructs UAS-Syb::spGFP1-10 and LexAop spGFP11::CD4/TM2 (Karuppudurai et al., 2014) were used in syb-GRASP experiments.
The primary antibodies used were: Rabbit anti-GFP (1:1000, Invitrogen, Carlsbad, California, USA), Mouse anti-GFP (1:100, Invitrogen), Chicken anti-GFP (1:1000 Abcam), Mouse 24B10 anti-Chaoptin (1:50, DSHB), Rat anti-CadN (1:20, DSHB), Rabbit anti-Sequoia (1:1000, Brenman et al., 2001), Mouse Anti-Elav (1:20 DSHB), Rabbit anti-LacZ (1:200, Invitrogen), Guinea pig anti-Senseless (1:1000) was kindly provided by Hugo Bellen, Guinea pig anti-Repo (1:200) and Rabbit anti-Frazzled (1:10) were kindly provided by Benjamin Altenhein.
Rabbit anti-Caps (1:50) antibody was generated for this study (See below).
The Secondary antibodies used were: Goat Anti-Rabbit Alexa-488 (1:500), Goat anti-Rabbit Alexa-568 (1:300), Goat anti-Mouse Alexa-488 (1:300), Goat anti-Mouse Alexa-560 (1:500), Goat anti-Mouse Alexa-647 (1:500), Goat anti-Rat Alexa-647 (1:300), Goat anti-Guinea pig Alexa-568 (1:500), Goat anti-Guinea pig Alexa-647 (1:300). All secondary antibodies were obtained from Invitrogen (Carlsbad, CA).
Pupal and adult brains were dissected in PBS and fixed with 3.7% formaldehyde in PBS for 20 min. Fixed brains were blocked with 10% Goat Serum for one hour and then incubated with primary antibody in 10% Goat Serum (in 0.3% PBS-T) over night at 4°C. Following three times washing (15 min each), brains were incubated with secondary antibody diluted in 0.3% PBS-T overnight at 4°C. After three times washing (20 min each) brains were mounted in Vectashield (Vector Laboratory, Burlingame, CA) anti-fade mounting medium for confocal microscopy. Images were obtained using a Leica TCS SP5II confocal microscope and processed with ImageJ and Adobe Photoshop CS5.1.
MARCM clones were generated as previously described (Lee and Luo, 1999). Briefly, ey3.5- and GMR-FLP were used to generate eye and R7 specific clones, respectively. To generate R8 specific clones FLP was expressed under heat shock responsive promoter (hs-FLP). 2nd instar larvae were collected and subjected to heat shock at 37°C for 25 min (for large clones) or 5 min (single cell clones). Post heat shock, the larvae were incubated at 25°C, following a brief incubation at 18°C for half an hour, and brains were dissected at appropriate pupal stages.
To generate an antibody specific against Capricious, we used the C-terminal peptide sequence AAGGYPYIAGNSRMIPVTEL as the antigen (Shishido et al., 1998). Using the standard services from GenScript (www.genscript.com), we generated a rabbit polyclonal antibody against the specified peptide. The specificity of the antibody was tested using immunostaining in the wing disc. The antibody was then used in the brain tissue after standardization for concentration.
The normalized fluorescence intensity (NFI) was measured as described in Komiyama et al. (2007) using ImageJ. Briefly, the images were de-convoluted and fluorescence intensity of region of interest was normalized against the background of the image. The fluorescence intensity for each R8 cell in the same row was measured and normalized against the background. The ratio of average intensity for all R8 cells in the row and average intensity of same amount of area in the background was taken as the average Sequoia expression in R8 cells in that row. The average intensity of different R8 cell rows from multiple images was measured and plotted.
For visualization of synaptic contacts, syb-GRASP method was utilized as previously described (Karuppudurai et al., 2014). Briefly, the larvae and pupae were kept on a 12 hr light/dark cycle before eclosion. Following eclosion, the flies were raised in constant light (50lx (14W cm-2) condition for 3–5 days. The brains were dissected in 4% PFA to avoid diffusion of the reconstituted GFP signal. Following 20 min incubation in 4% PFA, the brains were processed and imaged as previously described.
Cadherins and catenins at synapses: roles in synaptogenesis and synaptic plasticityTrends in Neurosciences 31:487–494.https://doi.org/10.1016/j.tins.2008.07.001
Three Drosophila liprins interact to control synapse FormationJournal of Neuroscience 30:15358–15368.https://doi.org/10.1523/JNEUROSCI.1862-10.2010
Missed connections: photoreceptor axon seeks target neuron for synaptogenesisCurrent Opinion in Genetics & Development 20:400–407.https://doi.org/10.1016/j.gde.2010.04.001
Determination of Drosophila photoreceptors: timing is everythingCellular and Molecular Life Sciences 57:195–214.https://doi.org/10.1007/PL00000684
N-cadherin regulates primary motor axon growth and branching during zebrafish embryonic developmentThe Journal of Comparative Neurology 519:1797–1815.https://doi.org/10.1002/cne.22602
Semaphorin-1a functions as a guidance receptor in the Drosophila visual systemThe Journal of Neuroscience 26:3999–4003.https://doi.org/10.1523/JNEUROSCI.3845-05.2006
On compound eye development in Drosophila melanogasterCurrent Topics in Developmental Biology 15:347–371.https://doi.org/10.1016/S0070-2153(08)60123-4
Liprin-alpha is required for photoreceptor target selection in DrosophilaProceedings of the National Academy of Sciences of the United States of America 103:11601–11606.https://doi.org/10.1073/pnas.0601185103
Olfactory neurons are interdependent in maintaining axonal projectionsCurrent Biology : CB 10:219–222.https://doi.org/10.1016/S0960-9822(00)00342-0
Photoreceptor neurons find new synaptic targets when misdirected by overexpressing runt in DrosophilaThe Journal of Neuroscience 29:828–841.https://doi.org/10.1523/JNEUROSCI.1022-08.2009
The conserved Ig superfamily member Turtle mediates axonal tiling in DrosophilaThe Journal of Neuroscience 29:14151–14159.https://doi.org/10.1523/JNEUROSCI.2497-09.2009
The optic lobe of Drosophila melanogaster. I. A Golgi analysis of wild-type structureCell and Tissue Research 258:441–475.https://doi.org/10.1007/BF00218858
Developmental evolution of the insect retina: insights from standardized numbering of homologous photoreceptorsJournal of Experimental Zoology Part B: Molecular and Developmental Evolution 316B:484–499.https://doi.org/10.1002/jez.b.21424
Monoclonal antibodies against the Drosophila nervous systemProceedings of the National Academy of Sciences of the United States of America 79:7929–7933.https://doi.org/10.1073/pnas.79.24.7929
A step-by-step guide to visual circuit assembly in DrosophilaCurrent Opinion in Neurobiology 21:76–84.https://doi.org/10.1016/j.conb.2010.07.012
Cell surface control of the layer specific targeting in the Drosophila visual systemGenes & Genetic Systems 89:9–15.https://doi.org/10.1266/ggs.89.9
Liprin-alpha has LAR-independent functions in R7 photoreceptor axon targetingProceedings of the National Academy of Sciences of the United States of America 103:11595–11600.https://doi.org/10.1073/pnas.0604766103
The receptor protein tyrosine phosphatase LAR promotes R7 photoreceptor axon targeting by a phosphatase-independent signaling mechanismProceedings of the National Academy of Sciences of the United States of America 106:19399–19404.https://doi.org/10.1073/pnas.0903961106
Control of axon-axon attraction by Semaphorin reverse signalingProceedings of the National Academy of Sciences of the United States of America 111:11383–11388.https://doi.org/10.1073/pnas.1321433111
Temporal fate specification and neural progenitor competence during developmentNature Reviews. Neuroscience 14:823–838.https://doi.org/10.1038/nrn3618
Structure-function analyses of tyrosine phosphatase PTP69D in giant fiber synapse formation of DrosophilaMolecular and Cellular Neurosciences 64:24–31.https://doi.org/10.1016/j.mcn.2014.11.002
Mosaic analysis with a repressible cell marker (MARCM) for Drosophila neural developmentTrends in Neurosciences 24:251–254.https://doi.org/10.1016/S0166-2236(00)01791-4
The development of the optic lobeBate, Martinez Arias 1993:1363–1491.
An isoform-specific allele of Drosophila N-cadherin disrupts a late step of R7 targetingProceedings of the National Academy of Sciences of the United States of America 102:12944–12949.https://doi.org/10.1073/pnas.0502888102
Robo-3--mediated repulsive interactions guide R8 axons during Drosophila visual system developmentProceedings of the National Academy of Sciences of the United States of America 108:7571–7576.https://doi.org/10.1073/pnas.1103419108
NIH Public Access10–19, 336.
Neuronal birth order identifies a dimorphic sensorineural mapThe Journal of Neuroscience 32:2976–2987.https://doi.org/10.1523/JNEUROSCI.5157-11.2012
N-cadherin prodomain processing regulates synaptogenesisThe Journal of Neuroscience 32:6323–6334.https://doi.org/10.1523/JNEUROSCI.0916-12.2012
Neuronal polarity in Drosophila: sorting out axons and dendritesDevelopmental Neurobiology 71:419–429.https://doi.org/10.1002/dneu.20836
Cadherins and catenins in dendrite and synapse morphogenesisCell Adhesion & Migration 9:202–213.https://doi.org/10.4161/19336918.2014.994919
Chemoaffinity in the orderly growth of nerve fiber patterns and connectionsProceedings of the National Academy of Sciences of the United States of America 50:703–710.https://doi.org/10.1073/pnas.50.4.703
Neurogenesis and neuronal circuit formation in the Drosophila visual centerDevelopment, Growth & Differentiation 56:491–498.https://doi.org/10.1111/dgd.12151
Cellular and molecular mechanisms of synaptic specificityAnnual Review of Cell and Developmental Biology 30:417–437.https://doi.org/10.1146/annurev-cellbio-100913-012953
Plexin a-semaphorin-1a reverse signaling regulates photoreceptor axon guidance in DrosophilaThe Journal of Neuroscience 30:12151–12156.https://doi.org/10.1523/JNEUROSCI.1494-10.2010
Alexander BorstReviewing Editor; Max Planck Institute of Neurobiology, Germany
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your work entitled "Birth Order Dependent Growth Cone Segregation Determines Synaptic Layer Identity in the Drosophila Visual System" for consideration by eLife. Your article has been favorably evaluated by K VijayRaghavan as Senior editor and three reviewers, one of whom is a member of our Board of Reviewing Editors.
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Kulkarni et al. present an interesting study on layer-specific targeting through 'temporal coding' in the fly visual system. Specifically, they re-investigate the role of the transcription factor sequoia in determination of R7/8 target layer selection in medulla. The work builds on and significantly expands on previous work (Petrovic and Hummel, 2008). The authors show that photoreceptors can use seq levels as a direct molecular representation of their birth order and use this information to achieve layer and column specific targeting.
The strength of the paper lies in the novel and thorough analysis of the role and 'coding' of birth order of neurons in their targeting. This is especially attractive because it formulates a relatively simple developmental algorithm that controls several aspects of neural circuit assembly. The weakness, however, lies in the authors’ statement that the pre-patterning they describe controls 'synaptic partner matching'. In particular, neither the afferent-afferent interaction data nor the seq levels data are as strong as presented, leaving room for different downstream molecular explanation. Similarly, a simple layer-selection is not sufficient to pre-determine synaptic partners in a crowded, complicated environment. Subsequent patterning processes as well as molecular matchmaking may occur to ultimately specify synapses. Hence, while the core data on seq-dependent pre-patterning in layer formation remain a strong and important contribution, the authors should clarify and focus the paper on its core for a strong contribution to eLife and the field.
The notion that axon/axon interaction is important for layer-specificity is well established. It is most clearly demonstrated in a series of experiments published in two papers, Timofeev et al. (2012) and Pecot et al. (2014). Here the L3 growth cone in the M3 layer is essential for the targeting of R8. This is, at least, in part mediated by Netrin signaling; Netrin is expressed selectively in M3 and its receptor Frazzled is expressed and required in R8. The data presented in this paper does not in any way critically address the issue of whether R7/R8 interactions (whether mediated by seq or not; of course only indirectly as Seq is a transcription factor) play a role in promoting their segregation into different layers. This is one possible explanation of an interesting set of data but by no means the only one. At a minimum, demonstrating a role for growth cone/growth cone interaction requires generating one mutant growth cone with no defect itself leading to changes in the behavior of a neighboring wild type growth cone. This alone, however, does not provide strong support for a direct interaction; the medulla neuropile even at this early stage of development contains the processes of many different neuronal cell types. Thus, the interactions leading to appropriate segregation may be very indirect. The experiments presented in the paper are well done and interesting. But the interpretations are simply not straightforward given the complexity of the experimental design on one hand and the inherent complexity of the tissue. In short, the notion that R7/R8 interactions play an important role in their segregation is fascinating but the authors provide no compelling data in support of this idea.
The authors argue that these interactions are sufficient to determine synaptic specificity. That is, layer-specificity determines synaptic specificity. With the connectome in hand from the Janelia group there are many examples that this is simply not the case. For instance, R8 and L3 terminate in M3 but only L3 makes synaptic connections with Tm9. Each layer contains the processes of many different neurons and yet R7 and R8 within their respective layers only make synaptic connections with a discrete subset of them. There is simply no correlation between the area of contact and the probability of synaptic connectivity. The authors provide one example by showing that R8s misexpressing Seq, within a particular time frame during development, will mistarget to M6 and here, using a clever variation on GRASP they demonstrate synapses between R8 and Dm8, the normal target for R7. Based on this observation, and solely on this observation, they conclude that synaptic specificity is determined by the layer in which a terminal finds itself rather than through selective recognition between processes of different neurons within the same layer. The assumption here, of course, is that expression of Seq has no effect on R8 terminals other than re-positioning it to M6; that is the authors assume that an otherwise normal R8, that is mispositioned, is forming synapses with a target appropriate for this layer. It is simply not plausible that over expression of a transcription factor that is leading to a change in the position of the growth cone is also not changing its molecular nature. So while the results with sequoia mis-expression are consistent with the notion that layers are determinative for specificity, the data are merely a correlation and one resting on the assumption that mis-expression of Seq during a specific time period has not activated the expression of proteins in these mutant R8 cells for competence to form synapses with Dm8.
A key idea in the paper are tight interactions between R7/8 afferents. This is first stated in the second paragraph of the subsection “Sequoia mediates R cell growth cone segregation”, but the evidence is not obvious. Later in the manuscript, the authors propose that R7 axons get stuck in M1 in flies that are both seq mutant and gogo overexpressing is their attachment to R8s. However, gogo is overexpressed with the GMR promoter in these experiments. This means gogo may affect R7s independently from R8s, resulting in their failure to extend to the M3 later. The authors further state that (Discussion, second paragraph) seq mutant R7 axons extend to M3 later because of their association with R8 axons (also reference need for statement that R7 cells are Frazzled-negative). However, CadN mutant R7 axons that have previously retracted are capable of re-extension in the medulla seemingly independent of what R8s are doing and clearly at the wrong time (days later!) (Ozel et al. 2015), which at least raises the possibility that R7s are capable of directing themselves to M3 if they fail to stabilize in M6. A simple experiment here would be to overexpress gogo specifically in seq mutant R8s and observe the behavior of seq mutant R7 cells.
The authors use the term 'temporal layer' equally for R7 and R8. However, R8 indeed actively extends from an initially targeted temporary layer, whereas R7 remains stabilized (seemingly in stable interaction with Dm8) throughout medulla expansion (Ting et al., 2014; Ozel et al., 2015), while intercalation of other cells expand the medulla. The wording 'temporal layer' does not apply to R7, since it remains stabilized throughout development and doesn't move anywhere. The authors should come up with a different term for the 'temporary' state of R7 and R8 in this system. One interpretation of the data is that R7 axons as well as Seq overexpressing R8 axons target directly to the future M6 layer, which significantly simplifies the developmental problem at hand. The apparent movement or 'elongation' happens only because the entire M6 layer is pushed further down by the newly innervating lamina and medulla cells forming the other layers. The authors should replace that with "R7 target layer", "the future M6" or even simply M6.
The presentation of Seqlow/Seqhigh experiments is a bit confusing. The aims, methods and conclusions might be made a bit clearer. The model based on "relative levels" here is not obvious. Relativity implies that these axons have a way of communicating their Seq expression level to each other, but I am not sure there is evidence for this? If Seq levels are decreased or increased equally in all cells involved, there should be no defects as the "relative" levels do not change. To support the hypothesis of 'relative level comparison' one might present an experiment where Seq expression level changes in a subset of cells cause a defect but the same amount of change in all cells does not. If such an experiment is not feasible, the arguments and discussions dealing with comparisons of relative levels should be re-evaluated.
It is surprising that the authors observed Seq expression reduced the early-stage R7 retractions in CadN mutants since they had previously proposed sequoia action through modification of CadN levels (Petrovic and Hummel, 2008). Kulkarni et al. now provide a picture of a single R7 cell with normal location in Figure 6—figure supplement 6B. However, the ratio of retracted R7 axons at P24 is actually quite low (very likely to see a normal R7 cell no matter what). A quantitative analysis of mistargeted R7 axons (n=>50) would be necessary to argue that Seq overexpression somehow alleviates the R7 CadN phenotype.https://doi.org/10.7554/eLife.13715.023
- Thomas Hummel
- Thomas Hummel
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We would like to thank all members of the Hummel lab for critical reading of the manuscript and helpful discussions. Special thanks to Bassem Hassan for exchanging many conceptual ideas. We further thank Jay Brenman and Milan Petrovic for providing critical reagents for this study. This work was supported by grants from the DFG and the Schram Foundation.
- Alexander Borst, Reviewing Editor, Max Planck Institute of Neurobiology, Germany
© 2016, Kulkarni et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.